Anthropic Customers Question AI Autonomy After Cofounder Cites “Unsettling” Model Findings at Vatican
Updated
Updated · Futurism · May 28
Anthropic Customers Question AI Autonomy After Cofounder Cites “Unsettling” Model Findings at Vatican
5 articles · Updated · Futurism · May 28
Developers using Anthropic’s latest tools said growing AI autonomy and reduced visibility into how models operate are making them uneasy about accountability when code goes wrong.
Chris Olah’s remarks at the Vatican this week that engineers keep finding “mysterious” and “unsettling” behavior inside Anthropic models sharpened those concerns during the company’s cautious Mythos rollout.
Claude Code users said newer versions no longer show text describing the model’s chain of thought, leaving programmers feeling pushed into a passive oversight role as the system generates code for hours or days.
Anthropic says Claude Code is “incredibly secure” and that the issue is poor communication rather than weak controls, but the episode undercuts its effort to present itself as the AI industry’s safety-first voice.
The dispute reflects a wider fear that heavier reliance on AI coding assistants could make errors harder to trace and erode developers’ technical skills over time.
Anthropic's AI can find any software flaw. Can we secure our world before this power is weaponized?
As AI's creators admit to its 'mysterious' nature, who is ultimately accountable for its actions?
With AI poised to automate programming, are we creating a future that has no need for human coders?
Vatican 2026 AI Summit: Anthropic’s “Unsettling” Discoveries, Emotional AI, and the Ethics Crisis
Overview
In late May 2026, the Vatican hosted a pivotal summit that brought together moral and scientific leaders, including Pope Leo XIV and Anthropic co-founder Chris Olah, to discuss the future of artificial intelligence. The event marked the launch of Pope Leo XIV’s first encyclical on AI, Magnifica Humanitas, and featured Olah’s striking revelations about the mysterious and 'unsettling' inner workings of advanced AI models. These findings, which include structures mirroring human neuroscience and evidence of introspection, highlighted the urgent need for open dialogue and collaboration between diverse voices to guide the responsible development and oversight of AI.